def run_kg(run_index):
# Store data for debugging
IS0 = pickle.load(open("enthalpy_N1_R3_Ukcal-mol", 'r'))
#IS0 = pickle.load(open("enthalpy_N3_R2_Ukcal-mol", 'r'))
# Generate the main object
sim = Optimizer()
# Assign simulation properties
#sim.hyperparameter_objective = MAP
sim.hyperparameter_objective = MLE
###################################################################################################
# File names
sim.fname_out = "enthalpy_kg.dat"
sim.fname_historical = "data_dumps/%d_reduced.history" % run_index
print "Waiting on %s to be written..." % sim.fname_historical,
while not os.path.exists(sim.fname_historical):
time.sleep(30)
print " DONE"
# Information sources, in order from expensive to cheap
sim.IS = [
lambda h, c, s: -1.0 * IS0[' '.join([''.join(h), c, s])],
]
sim.costs = [1.0]
sim.save_extra_files = True
sim.logger_fname = "data_dumps/%d_kg.log" % run_index
if os.path.exists(sim.logger_fname):
os.system("rm %s" % sim.logger_fname)
os.system("touch %s" % sim.logger_fname)
sim.obj_vs_cost_fname = "data_dumps/%d_kg.dat" % run_index
sim.mu_fname = "data_dumps/%d_mu_kg.dat" % run_index
sim.sig_fname = "data_dumps/%d_sig_kg.dat" % run_index
sim.sample_fname = "data_dumps/%d_sample_kg.dat" % run_index
sim.combos_fname = "data_dumps/%d_combos_kg.dat" % run_index
sim.hp_fname = "data_dumps/%d_hp_kg.dat" % run_index
sim.acquisition_fname = "data_dumps/%d_acq_kg.dat" % run_index
sim.historical_nsample = 10
########################################
sim.n_start = 20 # The number of starting MLE samples
sim.reopt = 10
sim.ramp_opt = None
sim.parallel = False
sim.acquisition = getNextSample_kg
# Possible compositions by default
sim.A = ["Cs", "MA", "FA"]
sim.B = ["Pb"]
sim.X = ["Cl", "Br", "I"]
sim.solvents = copy.deepcopy(solvents)
sim.S = list(set([v["name"] for k, v in sim.solvents.items()]))
sim.mixed_halides = True
sim.mixed_solvents = False
# Parameters for debugging and overwritting
sim.debug = False
sim.verbose = True
sim.overwrite = True # If True, warning, else Error
# Functional forms of our mean and covariance
# MEAN: 4 * mu_alpha + mu_zeta
# COV: sig_alpha * |X><X| + sig_beta * I_N + sig_zeta + MaternKernel(S, weights, sig_m)
SCALE = [2.0, 4.0][int(sim.mixed_halides)]
# _1, _2, _3 used as dummy entries
sim.mean = lambda _1, Y, theta: np.array(
[SCALE * theta.mu_alpha + theta.mu_zeta for _ in Y])
def cov(X, Y, theta):
A = theta.sig_alpha * np.dot(
np.array(X)[:, 1:-3],
np.array(X)[:, 1:-3].T)
B = theta.sig_beta * np.diag(np.ones(len(X)))
C = theta.sig_zeta
D = mk52(np.array(X)[:, -3:-1], [theta.l1, theta.l2], theta.sig_m)
return A + B + C + D
sim.cov = cov
sim.theta.bounds = {}
sim.theta.mu_alpha, sim.theta.bounds['mu_alpha'] = None, (
1E-3, lambda _, Y: max(Y))
sim.theta.sig_alpha, sim.theta.bounds['sig_alpha'] = None, (
1E-2, lambda _, Y: 10.0 * np.var(Y))
sim.theta.sig_beta, sim.theta.bounds['sig_beta'] = None, (
1E-2, lambda _, Y: 10.0 * np.var(Y))
sim.theta.mu_zeta, sim.theta.bounds['mu_zeta'] = None, (
1E-3, lambda _, Y: max(Y))
sim.theta.sig_zeta, sim.theta.bounds['sig_zeta'] = None, (
1E-2, lambda _, Y: 10.0 * np.var(Y))
sim.theta.sig_m, sim.theta.bounds['sig_m'] = None, (1E-2,
lambda _, Y: np.var(Y))
sim.theta.l1, sim.theta.bounds['l1'] = None, (1E-1, 1)
sim.theta.l2, sim.theta.bounds['l2'] = None, (1E-1, 1)
# NOTE! This is a reserved keyword in misoKG. We will generate a list of the same length
# of the information sources, and use this for scaling our IS.
# sim.theta.rho, sim.theta.bounds['rho'] = {"[0, 0]": 1}, (1E-1, 5.0)
# NOTE! This is a reserved keyword in misoKG. We will generate a list of the same length
# of the information sources, and use this for scaling our IS.
sim.theta.rho = {"[0, 0]": 1}
sim.theta.bounds['rho [0, 0]'] = (1, 1)
sim.theta.set_hp_names()
sim.primary_rho_opt = False
###################################################################################################
# Start simulation
sim.run()
def run(run_index,
model,
folder="data_dumps",
hp_opt="IS0",
sample_domain=1000):
'''
This function will run CO optimization using one of several coregionalization methods.
1. Pearson R Intrinsic Coregionalization Method (PRICM). This approach
will dynamically calculate the Pearson R value for the off-diagonals
in the ICM. Diagonals are kept as 1.0.
2. Intrinsic Coregionalization Method (ICM). This approach will use a
lower triangular matrix (L) of hyperparameters to generate the
coregionalization matrix B = LL^T.
Further, we can parameterize the hyperparameters in many ways:
1. IS0 - Only parameterize hyperparameters using values sampled at IS0.
2. Full - Parameterize hyperparameters using all sampled data.
3. Overlap - Parameterize hyperparameters using data that overlaps all IS.
**Parameters**
run_index: *int*
This is simply used for a naming convention.
model: *str*
The model to be used (PRICM or ICM).
folder: *str, optional*
What to name the folder where the data will go.
hp_opt: *str, optional*
With what data should the hyperparameters be parameterized.
Options: IS0, full, overlap
sample_domain: *int, optional*
How many data points to sample from the domain.
'''
hp_opt = hp_opt.lower()
allowed_hp_opt = ["is0", "full", "overlap"]
assert hp_opt in allowed_hp_opt, "Error, hp_opt (%s) not in %s" % (
hp_opt, ", ".join(allowed_hp_opt))
# Generate the main object
sim = Optimizer()
# Assign simulation properties
sim.hyperparameter_objective = MLE
###################################################################################################
# File names
sim.fname_out = None
sim.fname_historical = None
sim.logger_fname = "%s/%d_%s_%s.log" % (folder, run_index, model, hp_opt)
if os.path.exists(sim.logger_fname):
os.system("rm %s" % sim.logger_fname)
os.system("touch %s" % sim.logger_fname)
sim.obj_vs_cost_fname = None
sim.mu_fname = None
sim.sig_fname = None
sim.combos_fname = None
sim.hp_fname = None
sim.acquisition_fname = None
sim.save_extra_files = True
# Information sources, in order from expensive to cheap
rosenbrock = lambda x1, x2: (1.0 - x1)**2 + 100.0 * (x2 - x1**2)**2 - 456.3
sim.IS = [
lambda x1, x2: -1.0 * rosenbrock(x1, x2), lambda x1, x2: -1.0 *
(rosenbrock(x1, x2) + 10.0 * np.sin(10.0 * x1 + 5.0 * x2))
]
sim.costs = [1000.0, 1.0]
sim.save_extra_files = False
########################################
sim.numerical = True
sim.historical_nsample = 5
sim.domain = [(-2.0, 2.0), (-2.0, 2.0)]
sim.sample_n_from_domain = sample_domain
########################################
sim.n_start = 10 # The number of starting MLE samples
# sim.reopt = 20
sim.reopt = float('inf') # Never re-opt hyperparams
sim.ramp_opt = None
sim.parallel = False
# Parameters for debugging and overwritting
sim.debug = False
sim.verbose = True
sim.overwrite = True # If True, warning, else Error
sim.acquisition = getNextSample_misokg
# Functional forms of our mean and covariance
sim.mean = lambda X, Y, theta: np.array([-456.3 for _ in Y])
def cov_miso(X0, Y, theta, split=False):
Kx = squared(np.array(X0), [theta.l1], theta.sig_1)
Kx_l = squared(np.array(X0), [theta.l2], theta.sig_2)
return np.block([[Kx, Kx], [Kx, Kx + Kx_l]])
def cov_pricm(X0, Y, theta, split=False):
Kx = squared(np.array(X0), [theta.l1], theta.sig_1)
Kx = Kx + 1E-6 * np.eye(Kx.shape[0])
if model.lower() == "pricm":
Ks = np.array([
np.array([
theta.rho[str(sorted([i, j]))] for j in range(theta.n_IS)
]) for i in range(theta.n_IS)
])
elif model.lower() == "icm":
L = np.array([
np.array([
theta.rho[str(sorted([i, j]))] if i >= j else 0.0
for j in range(theta.n_IS)
]) for i in range(theta.n_IS)
])
# Force it to be positive semi-definite
Ks = L.dot(L.T)
if split:
return Ks, Kx
else:
return np.kron(Ks, Kx)
sim.theta.bounds = {}
sim.theta.sig_1, sim.theta.bounds['sig_1'] = None, (1E-2,
lambda _, Y: np.var(Y))
sim.theta.l1, sim.theta.bounds['l1'] = None, (1E-1, 1)
if model == "miso":
sim.cov = cov_miso
sim.theta.sig_2, sim.theta.bounds['sig_2'] = None, (
1E-2, lambda _, Y: np.var(Y))
sim.theta.l2, sim.theta.bounds['l2'] = None, (1E-1, 1)
sim.theta.rho = {
str(sorted([i, j])): 1.0
for i in range(len(sim.IS)) for j in range(i, len(sim.IS))
}
else:
sim.cov = cov_pricm
sim.theta.rho = {"[0, 0]": None, "[0, 1]": None, "[1, 1]": None}
if model.lower() == "icm":
sim.theta.rho = {
str(sorted([i, j])): None
for i in range(len(sim.IS)) for j in range(i, len(sim.IS))
}
elif model.lower() == "pricm":
sim.theta.rho = {
str(sorted([i, j])): 1.0
for i in range(len(sim.IS)) for j in range(i, len(sim.IS))
}
sim.dynamic_pc = True
else:
raise Exception("Invalid model. Use MISO, ICM, or PRICM")
for k in sim.theta.rho.keys():
sim.theta.bounds['rho %s' % k] = (0.1, 1.0)
a, b = eval(k)
if a != b:
sim.theta.bounds['rho %s' % k] = (0.01, 1.0 - 1E-6)
sim.theta.set_hp_names()
# Define how we update hyperparameters
hp_opt = hp_opt.lower()
if hp_opt == "is0":
sim.update_hp_only_with_IS0 = True
sim.update_hp_only_with_overlapped = False
elif hp_opt == "overlap":
sim.update_hp_only_with_IS0 = False
sim.update_hp_only_with_overlapped = True
elif hp_opt == "full":
sim.update_hp_only_with_IS0 = False
sim.update_hp_only_with_overlapped = False
else:
raise Exception("Unknown hp_opt (%s)." % hp_opt)
# These should be False by default, but ensure they are
sim.theta.normalize_L = False
sim.theta.normalize_Ks = False
sim.preconditioned = False
# Assign our likelihood function.
sim.loglike = g_loglike
###################################################################################################
# Start simulation
sim.iteration_kill_switch = None
sim.cost_kill_switch = 100000
sim.run()
def run_misokg(run_index):
# Store data for debugging
IS0 = pickle.load(open("enthalpy_N1_R3_Ukcal-mol", 'r'))
IS1 = pickle.load(open("enthalpy_N1_R2_Ukcal-mol", 'r'))
# Generate the main object
sim = Optimizer()
# Assign simulation properties
#sim.hyperparameter_objective = MAP
sim.hyperparameter_objective = MLE
###################################################################################################
# File names
sim.fname_out = "enthalpy_misokg.dat"
sim.fname_historical = None
# Information sources, in order from expensive to cheap
sim.IS = [
lambda h, c, s: -1.0 * IS0[' '.join([''.join(h), c, s])],
lambda h, c, s: -1.0 * IS1[' '.join([''.join(h), c, s])]
]
sim.costs = [1.0, 0.1]
sim.logger_fname = "data_dumps/%d_misokg.log" % run_index
if os.path.exists(sim.logger_fname):
os.system("rm %s" % sim.logger_fname)
os.system("touch %s" % sim.logger_fname)
sim.obj_vs_cost_fname = "data_dumps/%d_misokg.dat" % run_index
sim.mu_fname = "data_dumps/%d_mu_misokg.dat" % run_index
sim.sig_fname = "data_dumps/%d_sig_misokg.dat" % run_index
sim.combos_fname = "data_dumps/%d_combos_misokg.dat" % run_index
sim.hp_fname = "data_dumps/%d_hp_misokg.dat" % run_index
sim.acquisition_fname = "data_dumps/%d_acq_misokg.dat" % run_index
sim.save_extra_files = True
########################################
# Override the possible combinations with the reduced list of IS0
# Because we do this, we should also generate our own historical sample
combos_no_IS = [
k[1] + "Pb" + k[0] + "_" + k[2]
for k in [key.split() for key in IS0.keys()]
]
sim.historical_nsample = 10
choices = np.random.choice(combos_no_IS,
sim.historical_nsample,
replace=False)
tmp_data = pal_strings.alphaToNum(choices,
solvents,
mixed_halides=True,
name_has_IS=False)
data = []
for IS in range(len(sim.IS)):
for i, d in enumerate(tmp_data):
h, c, _, s, _ = pal_strings.parseName(pal_strings.parseNum(
d, solvents, mixed_halides=True, num_has_IS=False),
name_has_IS=False)
c = c[0]
data.append([IS] + d + [sim.IS[IS](h, c, s)])
sim.fname_historical = "data_dumps/%d.history" % run_index
pickle.dump(data, open(sim.fname_historical, 'w'))
simple_data = [d for d in data if d[0] == 0]
pickle.dump(simple_data,
open("data_dumps/%d_reduced.history" % run_index, 'w'))
########################################
sim.n_start = 10 # The number of starting MLE samples
sim.reopt = 20
sim.ramp_opt = None
sim.parallel = False
# Possible compositions by default
sim.A = ["Cs", "MA", "FA"]
sim.B = ["Pb"]
sim.X = ["Cl", "Br", "I"]
sim.solvents = copy.deepcopy(solvents)
sim.S = list(set([v["name"] for k, v in sim.solvents.items()]))
sim.mixed_halides = True
sim.mixed_solvents = False
# Parameters for debugging and overwritting
sim.debug = False
sim.verbose = True
sim.overwrite = True # If True, warning, else Error
sim.acquisition = getNextSample_misokg
# Functional forms of our mean and covariance
# MEAN: 4 * mu_alpha + mu_zeta
# COV: sig_alpha * |X><X| + sig_beta * I_N + sig_zeta + MaternKernel(S, weights, sig_m)
SCALE = [2.0, 4.0][int(sim.mixed_halides)]
# _1, _2, _3 used as dummy entries
def mean(X, Y, theta):
mu = np.array([SCALE * theta.mu_alpha + theta.mu_zeta for _ in Y])
return mu
sim.mean = mean
def cov_old(X, Y, theta):
A = theta.sig_alpha * np.dot(
np.array(X)[:, 1:-3],
np.array(X)[:, 1:-3].T)
B = theta.sig_beta * np.diag(np.ones(len(X)))
C = theta.sig_zeta
D = mk52(np.array(X)[:, -3:-1], [theta.l1, theta.l2], theta.sig_m)
return theta.rho_matrix(X) * (A + B + C + D)
def cov(X0, Y, theta):
A = theta.sig_alpha * np.dot(
np.array(X0)[:, :-3],
np.array(X0)[:, :-3].T)
B = theta.sig_beta * np.diag(np.ones(len(X0)))
C = theta.sig_zeta
D = mk52(np.array(X0)[:, -3:-1], [theta.l1, theta.l2], theta.sig_m)
Kx = A + B + C + D
Ks = np.array([
np.array(
[theta.rho[str(sorted([i, j]))] for j in range(theta.n_IS)])
for i in range(theta.n_IS)
])
if theta.normalize_Ks:
Ks = Ks / np.linalg.norm(Ks)
e = np.diag(np.array([theta.e1, theta.e2]))
Ks = e.dot(Ks.dot(e))
return np.kron(Ks, Kx)
sim.cov = cov
sim.theta.bounds = {}
sim.theta.mu_alpha, sim.theta.bounds['mu_alpha'] = None, (
1E-3, lambda _, Y: max(Y))
sim.theta.sig_alpha, sim.theta.bounds['sig_alpha'] = None, (
1E-2, lambda _, Y: 10.0 * np.var(Y))
sim.theta.sig_beta, sim.theta.bounds['sig_beta'] = None, (
1E-2, lambda _, Y: 10.0 * np.var(Y))
sim.theta.mu_zeta, sim.theta.bounds['mu_zeta'] = None, (
1E-3, lambda _, Y: max(Y))
sim.theta.sig_zeta, sim.theta.bounds['sig_zeta'] = None, (
1E-2, lambda _, Y: 10.0 * np.var(Y))
sim.theta.sig_m, sim.theta.bounds['sig_m'] = None, (1E-2,
lambda _, Y: np.var(Y))
sim.theta.l1, sim.theta.bounds['l1'] = None, (1E-1, 1)
sim.theta.l2, sim.theta.bounds['l2'] = None, (1E-1, 1)
sim.theta.e1, sim.theta.bounds['e1'] = None, (1E-1, 1.0)
sim.theta.e2, sim.theta.bounds['e2'] = None, (1E-1, 1.0)
# # NOTE! This is a reserved keyword in misoKG. We will generate a list of the same length
# # of the information sources, and use this for scaling our IS.
sim.theta.rho = {"[0, 0]": 1.0, "[0, 1]": 0.96, "[1, 1]": 1.0}
sim.theta.bounds['rho [0, 0]'] = (0.1, 1.0)
sim.theta.bounds['rho [0, 1]'] = (0.1, 1.0)
sim.theta.bounds['rho [1, 1]'] = (0.1, 1.0)
sim.theta.set_hp_names()
sim.primary_rho_opt = False
sim.update_hp_only_with_IS0 = False
sim.update_hp_only_with_overlapped = False
sim.theta.normalize_L = False
sim.theta.normalize_Ks = False
# This was a test feature that actually over-wrote rho to be PSD
# sim.force_rho_psd = True
sim.recommendation_kill_switch = "FAPbBrBrCl_THTO_0"
###################################################################################################
# Start simulation
sim.run()
def run(run_index, folder="data_dumps", sample_domain=1000):
'''
This function will run CO optimization using one of several coregionalization methods.
1. Pearson R Intrinsic Coregionalization Method (PRICM). This approach
will dynamically calculate the Pearson R value for the off-diagonals
in the ICM. Diagonals are kept as 1.0.
2. Intrinsic Coregionalization Method (ICM). This approach will use a
lower triangular matrix (L) of hyperparameters to generate the
coregionalization matrix B = LL^T.
Further, we can parameterize the hyperparameters in many ways:
1. IS0 - Only parameterize hyperparameters using values sampled at IS0.
2. Full - Parameterize hyperparameters using all sampled data.
3. Overlap - Parameterize hyperparameters using data that overlaps all IS.
**Parameters**
run_index: *int*
This is simply used for a naming convention.
folder: *str, optional*
What to name the folder where the data will go.
sample_domain: *int, optional*
How many data points to sample from the domain.
'''
# Generate the main object
sim = Optimizer()
# Assign simulation properties
sim.hyperparameter_objective = MLE
sim.acquisition = getNextSample_EI
###################################################################################################
# File names
sim.fname_out = None
sim.fname_historical = None
sim.logger_fname = "%s/%d_%s.log" % (folder, run_index, "ei")
if os.path.exists(sim.logger_fname):
os.system("rm %s" % sim.logger_fname)
os.system("touch %s" % sim.logger_fname)
sim.obj_vs_cost_fname = None
sim.mu_fname = None
sim.sig_fname = None
sim.combos_fname = None
sim.hp_fname = None
sim.acquisition_fname = None
sim.save_extra_files = True
# Information sources, in order from expensive to cheap
IS0 = pickle.load(open("IS0.pickle", 'r'))
sim.IS = [
lambda x1: -1.0 * IS0[int((x1 - 0.5) * 1000.0)][0],
]
sim.costs = [
np.mean([IS[1] for IS in IS0]),
]
sim.save_extra_files = False
########################################
sim.numerical = True
sim.historical_nsample = 5
sim.domain = [(0.5, 2.5)]
sim.sample_n_from_domain = sample_domain
########################################
sim.n_start = 10 # The number of starting MLE samples
# sim.reopt = 20
sim.reopt = float("inf")
sim.ramp_opt = None
sim.parallel = False
# Parameters for debugging and overwritting
sim.debug = False
sim.verbose = True
sim.overwrite = True # If True, warning, else Error
# Functional forms of our mean and covariance
sim.mean = lambda X, Y, theta: np.array([0.0 for _ in Y])
def cov(X0, Y, theta):
return squared(np.array(X0), [theta.l1], theta.sig_1)
sim.cov = cov
sim.theta.bounds = {}
sim.theta.sig_1, sim.theta.bounds['sig_1'] = None, (1E-2,
lambda _, Y: np.var(Y))
sim.theta.l1, sim.theta.bounds['l1'] = None, (1E-1, 1)
sim.theta.rho = {
str(sorted([i, j])): 1.0
for i in range(len(sim.IS)) for j in range(i, len(sim.IS))
}
for k in sim.theta.rho.keys():
sim.theta.bounds['rho %s' % k] = (0.1, 1.0)
a, b = eval(k)
if a != b:
sim.theta.bounds['rho %s' % k] = (0.01, 1.0 - 1E-6)
sim.theta.set_hp_names()
# Define how we update hyperparameters
sim.update_hp_only_with_IS0 = False
sim.update_hp_only_with_overlapped = False
# These should be False by default, but ensure they are
sim.theta.normalize_L = False
sim.theta.normalize_Ks = False
sim.preconditioned = False
# Assign our likelihood function.
sim.loglike = g_loglike
###################################################################################################
# Start simulation
sim.iteration_kill_switch = None
sim.cost_kill_switch = 3000
sim.run()
def run_ei(run_index, SAMPLE_DOMAIN=1000):
FOLDER = "RNS%d" % SAMPLE_DOMAIN
sffx = "ei"
# Generate the main object
sim = Optimizer()
# Assign simulation properties
#if use_MAP:
# sim.hyperparameter_objective = MAP
#else:
sim.hyperparameter_objective = MLE
###################################################################################################
# File names
sim.fname_out = None
sim.fname_historical = None
sim.logger_fname = "%s/%d_%s.log" % (FOLDER, run_index, sffx)
if os.path.exists(sim.logger_fname):
os.system("rm %s" % sim.logger_fname)
os.system("touch %s" % sim.logger_fname)
sim.obj_vs_cost_fname = None
sim.mu_fname = None
sim.sig_fname = None
sim.combos_fname = None
sim.hp_fname = None
sim.acquisition_fname = None
sim.save_extra_files = True
# Information sources, in order from expensive to cheap
rosenbrock = lambda x1, x2: (1.0 - x1)**2 + 100.0 * (x2 - x1**2)**2 - 456.3
sim.IS = [lambda x1, x2: -1.0 * rosenbrock(x1, x2)]
sim.costs = [1000.0]
########################################
sim.numerical = True
sim.historical_nsample = 5
sim.domain = [(-2.0, 2.0), (-2.0, 2.0)]
sim.sample_n_from_domain = SAMPLE_DOMAIN
########################################
sim.n_start = 10 # The number of starting MLE samples
sim.reopt = 20
sim.ramp_opt = None
sim.parallel = False
# Parameters for debugging and overwritting
sim.debug = False
sim.verbose = True
sim.overwrite = True # If True, warning, else Error
# Functional forms of our mean and covariance
sim.mean = lambda X, Y, theta: np.array([-456.3 for _ in Y])
def cov(X0, Y, theta):
return squared(np.array(X0)[:, 1:], [theta.l1, theta.l2], theta.sig_1)
sim.cov = cov
sim.theta.bounds = {}
sim.theta.sig_1, sim.theta.bounds['sig_1'] = None, (1E-2,
lambda _, Y: np.var(Y))
sim.theta.l1, sim.theta.bounds['l1'] = None, (1E-1, 1)
sim.theta.l2, sim.theta.bounds['l2'] = None, (1E-1, 1)
sim.theta.rho = {
str(sorted([i, j])): 1.0
for i in range(len(sim.IS)) for j in range(i, len(sim.IS))
}
for k in sim.theta.rho.keys():
sim.theta.bounds['rho %s' % k] = (0.1, 1.0)
a, b = eval(k)
if a != b:
sim.theta.bounds['rho %s' % k] = (0.01, 1.0 - 1E-6)
sim.theta.set_hp_names()
sim.update_hp_only_with_IS0 = False
sim.update_hp_only_with_overlapped = False
sim.theta.normalize_L = False
sim.theta.normalize_Ks = False
###################################################################################################
# Start simulation
sim.iteration_kill_switch = 200
sim.cost_kill_switch = 10000
sim.run()
def run_misokg(run_index,
sffx="misokg",
scaled=False,
loose=False,
very_loose=False,
use_MAP=False,
upper=1.0,
run_unitary=False,
use_miso=False):
SAMPLE_DOMAIN = 1000
# Generate the main object
sim = Optimizer()
# Assign simulation properties
if use_MAP:
sim.hyperparameter_objective = MAP
else:
sim.hyperparameter_objective = MLE
###################################################################################################
# File names
sim.fname_out = None
sim.fname_historical = None
sim.logger_fname = "data_dumps/%d_%s.log" % (run_index, sffx)
if os.path.exists(sim.logger_fname):
os.system("rm %s" % sim.logger_fname)
os.system("touch %s" % sim.logger_fname)
sim.obj_vs_cost_fname = None
sim.mu_fname = None
sim.sig_fname = None
sim.combos_fname = None
#sim.hp_fname = None
sim.hp_fname = "data_dumps/%d_HP_%s.log" % (run_index, sffx)
sim.acquisition_fname = None
sim.save_extra_files = True
# Information sources, in order from expensive to cheap
rosenbrock = lambda x1, x2: (1.0 - x1)**2 + 100.0 * (x2 - x1**2)**2 - 456.3
sim.IS = [
lambda x1, x2: -1.0 * rosenbrock(x1, x2), lambda x1, x2: -1.0 *
(rosenbrock(x1, x2) + 0.1 * np.sin(10.0 * x1 + 5.0 * x2))
]
#sim.IS = [
# lambda x1, x2: (1.0 - x1)**2 + 100.0 * (x2 - x1**2)**2 - 456.3 + np.random.normal()
# lambda x1, x2: (1.0 - x1)**2 + 100.0 * (x2 - x1**2)**2 - 456.3 + 2.0 * np.sin(10.0 * x1 + 5.0 * x2)
#]
sim.costs = [1000.0, 1.0]
########################################
sim.numerical = True
sim.historical_nsample = 5
sim.domain = [(-2.0, 2.0), (-2.0, 2.0)]
sim.sample_n_from_domain = SAMPLE_DOMAIN
########################################
sim.n_start = 10 # The number of starting MLE samples
sim.reopt = 20
sim.ramp_opt = None
sim.parallel = False
# Parameters for debugging and overwritting
sim.debug = False
sim.verbose = True
sim.overwrite = True # If True, warning, else Error
sim.acquisition = getNextSample_misokg
# Functional forms of our mean and covariance
sim.mean = lambda X, Y, theta: np.array([-456.3 for _ in Y])
def cov_miso(X0, Y, theta):
Kx = squared(np.array(X0)[:, 1:], [theta.l1, theta.l2], theta.sig_1)
Kx_l = squared(np.array(X0)[:, 1:], [theta.l3, theta.l4], theta.sig_2)
return np.block([[Kx, Kx], [Kx, Kx + Kx_l]])
def cov_bonilla(X0, Y, theta):
#Kx = mk52(np.array(X0)[:, 1:], [theta.l1, theta.l2], theta.sig_m)
Kx = squared(np.array(X0)[:, 1:], [theta.l1, theta.l2], theta.sig_m)
Kx = Kx + 1E-6 * np.eye(Kx.shape[0])
if run_unitary:
Ks = np.array([[1.0, 1.0 - 1E-6], [1.0 - 1E-6, 1.0]])
else:
L = np.array([
np.array([
theta.rho[str(sorted([i, j]))] if i >= j else 0.0
for j in range(theta.n_IS)
]) # Lower triangulary
for i in range(theta.n_IS)
])
Ks = L.dot(L.T)
e = np.diag(np.array([theta.e1, theta.e2]))
Ks = np.matmul(e, np.matmul(Ks, e))
K = np.kron(Ks, Kx)
return np.kron(Ks, Kx)
if use_miso:
sim.cov = cov_miso
sim.theta.bounds = {}
sim.theta.sig_1, sim.theta.bounds['sig_1'] = None, (
1E-2, lambda _, Y: np.var(Y))
sim.theta.sig_2, sim.theta.bounds['sig_2'] = None, (
1E-2, lambda _, Y: np.var(Y))
sim.theta.l1, sim.theta.bounds['l1'] = None, (1E-1, 1)
sim.theta.l2, sim.theta.bounds['l2'] = None, (1E-1, 1)
sim.theta.l3, sim.theta.bounds['l3'] = None, (1E-1, 1)
sim.theta.l4, sim.theta.bounds['l4'] = None, (1E-1, 1)
sim.theta.rho = {"[0, 0]": 1.0, "[0, 1]": 1.0, "[1, 1]": 1.0}
sim.theta.bounds['rho [0, 0]'] = (1.0, 1.0)
sim.theta.bounds['rho [0, 1]'] = (1.0, 1.0)
sim.theta.bounds['rho [1, 1]'] = (1.0, 1.0)
else:
sim.cov = cov_bonilla
sim.theta.bounds = {}
sim.theta.sig_m, sim.theta.bounds['sig_m'] = None, (
1E-2, lambda _, Y: np.var(Y))
sim.theta.l1, sim.theta.bounds['l1'] = None, (1E-1, 1)
sim.theta.l2, sim.theta.bounds['l2'] = None, (1E-1, 1)
if scaled:
sim.theta.e1, sim.theta.bounds['e1'] = None, (1E-1, upper)
sim.theta.e2, sim.theta.bounds['e2'] = None, (1E-1, upper)
else:
sim.theta.e1, sim.theta.bounds['e1'] = 1.0, (1E-1, upper)
sim.theta.e2, sim.theta.bounds['e2'] = 1.0, (1E-1, upper)
# NOTE! This is a reserved keyword in misoKG. We will generate a list of the same length
# of the information sources, and use this for scaling our IS.
if very_loose:
sim.theta.rho = {"[0, 0]": None, "[0, 1]": None, "[1, 1]": None}
elif loose:
sim.theta.rho = {"[0, 0]": 1.0, "[0, 1]": None, "[1, 1]": 1.0}
elif run_unitary:
sim.theta.rho = {"[0, 0]": 1.0, "[0, 1]": 1.0, "[1, 1]": 1.0}
else:
raise Exception("What is trying to be run?")
sim.theta.bounds['rho [0, 0]'] = (0.1, upper)
sim.theta.bounds['rho [0, 1]'] = (0.1, upper)
sim.theta.bounds['rho [1, 1]'] = (0.01, upper)
sim.theta.set_hp_names()
sim.update_hp_only_with_IS0 = False
sim.update_hp_only_with_overlapped = False
sim.theta.normalize_L = False
sim.theta.normalize_Ks = False
###################################################################################################
# Start simulation
sim.iteration_kill_switch = 200
sim.cost_kill_switch = 10000
sim.run()
def run_misokg(run_index, sffx="misokg", SAMPLE_DOMAIN=1000):
FOLDER = "RNS%d" % SAMPLE_DOMAIN
scaled = False
dpc = False
invert_dpc = False
scaled = False
use_I = False
use_J = False
use_miso = False
if sffx == "misokg":
use_miso = True
elif sffx == "bdpc":
dpc = True
elif sffx == "bidpc":
dpc = True
invert_dpc = True
elif sffx == "bvl":
pass
elif sffx == "bsvl":
scaled = True
elif sffx == "bI":
use_I = True
elif sffx == "bu":
use_J = True
else:
raise Exception("This sffx (%s) is not accounted for." % sffx)
# Generate the main object
sim = Optimizer()
# Assign simulation properties
#if use_MAP:
# sim.hyperparameter_objective = MAP
#else:
sim.hyperparameter_objective = MLE
###################################################################################################
# File names
sim.fname_out = None
sim.fname_historical = None
sim.logger_fname = "%s/%d_%s.log" % (FOLDER, run_index, sffx)
if os.path.exists(sim.logger_fname):
os.system("rm %s" % sim.logger_fname)
os.system("touch %s" % sim.logger_fname)
sim.obj_vs_cost_fname = None
sim.mu_fname = None
sim.sig_fname = None
sim.combos_fname = None
sim.hp_fname = None
sim.acquisition_fname = None
sim.save_extra_files = True
# Information sources, in order from expensive to cheap
rosenbrock = lambda x1, x2: (1.0 - x1)**2 + 100.0 * (x2 - x1**2)**2 - 456.3
sim.IS = [
lambda x1, x2: -1.0 * rosenbrock(x1, x2), lambda x1, x2: -1.0 *
(rosenbrock(x1, x2) + 0.1 * np.sin(10.0 * x1 + 5.0 * x2))
]
#sim.IS = [
# lambda x1, x2: (1.0 - x1)**2 + 100.0 * (x2 - x1**2)**2 - 456.3 + np.random.normal()
# lambda x1, x2: (1.0 - x1)**2 + 100.0 * (x2 - x1**2)**2 - 456.3 + 2.0 * np.sin(10.0 * x1 + 5.0 * x2)
#]
sim.costs = [1000.0, 1.0]
########################################
sim.numerical = True
sim.historical_nsample = 5
sim.domain = [(-2.0, 2.0), (-2.0, 2.0)]
sim.sample_n_from_domain = SAMPLE_DOMAIN
########################################
sim.n_start = 10 # The number of starting MLE samples
sim.reopt = 20
sim.ramp_opt = None
sim.parallel = False
# Parameters for debugging and overwritting
sim.debug = False
sim.verbose = True
sim.overwrite = True # If True, warning, else Error
sim.acquisition = getNextSample_misokg
# Functional forms of our mean and covariance
sim.mean = lambda X, Y, theta: np.array([-456.3 for _ in Y])
def cov_miso(X0, Y, theta):
Kx = squared(np.array(X0)[:, 1:], [theta.l1, theta.l2], theta.sig_1)
Kx_l = squared(np.array(X0)[:, 1:], [theta.l3, theta.l4], theta.sig_2)
return np.block([[Kx, Kx], [Kx, Kx + Kx_l]])
def cov_bonilla(X0, Y, theta):
Kx = squared(np.array(X0)[:, 1:], [theta.l1, theta.l2], theta.sig_1)
Kx = Kx + 1E-6 * np.eye(Kx.shape[0])
if use_J:
Ks = np.ones((theta.n_IS, theta.n_IS)) * (1.0 - 1E-6) + np.eye(
theta.n_IS) * 1E-6
elif use_I:
Ks = np.eye(theta.n_IS)
elif dpc and invert_dpc:
Ks = np.array([
np.array([
1.0 if i != j else theta.rho["[0, %d]" % i]**(-2.0)
for j in range(theta.n_IS)
]) for i in range(theta.n_IS)
])
elif dpc:
Ks = np.array([
np.array([
theta.rho[str(sorted([i, j]))] for j in range(theta.n_IS)
]) for i in range(theta.n_IS)
])
else:
L = np.array([
np.array([
theta.rho[str(sorted([i, j]))] if i >= j else 0.0
for j in range(theta.n_IS)
]) for i in range(theta.n_IS)
])
# Force it to be positive semi-definite
Ks = L.dot(L.T)
if theta.n_IS == 2:
e = np.diag(np.array([theta.e1, theta.e2]))
elif theta.n_IS == 3:
e = np.diag(np.array([theta.e1, theta.e2, theta.e3]))
else:
raise Exception("HOW?")
Ks = e.dot(Ks.dot(e))
return np.kron(Ks, Kx)
sim.theta.bounds = {}
sim.theta.sig_1, sim.theta.bounds['sig_1'] = None, (1E-2,
lambda _, Y: np.var(Y))
sim.theta.l1, sim.theta.bounds['l1'] = None, (1E-1, 1)
sim.theta.l2, sim.theta.bounds['l2'] = None, (1E-1, 1)
if use_miso:
sim.cov = cov_miso
sim.theta.sig_2, sim.theta.bounds['sig_2'] = None, (
1E-2, lambda _, Y: np.var(Y))
sim.theta.l3, sim.theta.bounds['l3'] = None, (1E-1, 1)
sim.theta.l4, sim.theta.bounds['l4'] = None, (1E-1, 1)
sim.theta.rho = {
str(sorted([i, j])): 1.0
for i in range(len(sim.IS)) for j in range(i, len(sim.IS))
}
else:
sim.cov = cov_bonilla
if scaled:
sim.theta.e1, sim.theta.bounds['e1'] = None, (1E-1, 1.0)
sim.theta.e2, sim.theta.bounds['e2'] = None, (1E-1, 1.0)
else:
sim.theta.e1, sim.theta.bounds['e1'] = 1.0, (1E-1, 1.0)
sim.theta.e2, sim.theta.bounds['e2'] = 1.0, (1E-1, 1.0)
sim.theta.rho = {"[0, 0]": None, "[0, 1]": None, "[1, 1]": None}
if dpc or use_I or use_J:
sim.theta.rho = {
str(sorted([i, j])): 1.0
for i in range(len(sim.IS)) for j in range(i, len(sim.IS))
}
sim.dynamic_pc = dpc
for k in sim.theta.rho.keys():
sim.theta.bounds['rho %s' % k] = (0.1, 1.0)
a, b = eval(k)
if a != b:
sim.theta.bounds['rho %s' % k] = (0.01, 1.0 - 1E-6)
sim.theta.set_hp_names()
sim.update_hp_only_with_IS0 = False
sim.update_hp_only_with_overlapped = False
sim.theta.normalize_L = False
sim.theta.normalize_Ks = False
###################################################################################################
# Start simulation
sim.iteration_kill_switch = 200
sim.cost_kill_switch = 10000
#sim.cost_kill_switch = sim.iteration_kill_switch * sim.costs[0]
sim.run()
def run_misokg(run_index):
# Store data for debugging
IS0 = pickle.load(open("enthalpy_N1_R3_Ukcal-mol", 'r'))
IS1 = pickle.load(open("enthalpy_N1_R2_Ukcal-mol", 'r'))
# Generate the main object
sim = Optimizer()
# Assign simulation properties
#sim.hyperparameter_objective = MAP
sim.hyperparameter_objective = MLE
###################################################################################################
# File names
sim.fname_out = "enthalpy_misokg.dat"
sim.fname_historical = None
# Information sources, in order from expensive to cheap
sim.IS = [
lambda h, c, s: -1.0 * IS0[' '.join([''.join(h), c, s])],
lambda h, c, s: -1.0 * IS1[' '.join([''.join(h), c, s])]
]
sim.costs = [
1.0,
0.1,
]
sim.logger_fname = "data_dumps/%d_misokg.log" % run_index
if os.path.exists(sim.logger_fname):
os.system("rm %s" % sim.logger_fname)
os.system("touch %s" % sim.logger_fname)
sim.obj_vs_cost_fname = "data_dumps/%d_misokg.dat" % run_index
sim.mu_fname = "data_dumps/%d_mu_misokg.dat" % run_index
sim.sig_fname = "data_dumps/%d_sig_misokg.dat" % run_index
sim.combos_fname = "data_dumps/%d_combos_misokg.dat" % run_index
sim.hp_fname = "data_dumps/%d_hp_misokg.dat" % run_index
sim.acquisition_fname = "data_dumps/%d_acq_misokg.dat" % run_index
sim.save_extra_files = True
########################################
# Override the possible combinations with the reduced list of IS0
# Because we do this, we should also generate our own historical sample
combos_no_IS = [
k[1] + "Pb" + k[0] + "_" + k[2]
for k in [key.split() for key in IS0.keys()]
]
#sim.historical_nsample = 240
sim.historical_nsample = 10
choices = np.random.choice(combos_no_IS,
sim.historical_nsample,
replace=False)
tmp_data = pal_strings.alphaToNum(choices,
solvents,
mixed_halides=True,
name_has_IS=False)
data = []
for IS in range(len(sim.IS)):
for i, d in enumerate(tmp_data):
h, c, _, s, _ = pal_strings.parseName(pal_strings.parseNum(
d, solvents, mixed_halides=True, num_has_IS=False),
name_has_IS=False)
c = c[0]
data.append([IS] + d + [sim.IS[IS](h, c, s)])
sim.fname_historical = "data_dumps/%d.history" % run_index
pickle.dump(data, open(sim.fname_historical, 'w'))
simple_data = [d for d in data if d[0] == 0]
pickle.dump(simple_data,
open("data_dumps/%d_reduced.history" % run_index, 'w'))
########################################
sim.n_start = 10 # The number of starting MLE samples
sim.reopt = 10
sim.ramp_opt = None
sim.parallel = False
# Possible compositions by default
sim.A = ["Cs", "MA", "FA"]
sim.B = ["Pb"]
sim.X = ["Cl", "Br", "I"]
sim.solvents = copy.deepcopy(solvents)
sim.S = list(set([v["name"] for k, v in sim.solvents.items()]))
sim.mixed_halides = True
sim.mixed_solvents = False
# Parameters for debugging and overwritting
sim.debug = False
sim.verbose = True
sim.overwrite = True # If True, warning, else Error
sim.acquisition = getNextSample_misokg
# Functional forms of our mean and covariance
# MEAN: 4 * mu_alpha + mu_zeta
# COV: sig_alpha * |X><X| + sig_beta * I_N + sig_zeta + MaternKernel(S, weights, sig_m)
SCALE = [2.0, 4.0][int(sim.mixed_halides)]
# _1, _2, _3 used as dummy entries
def mean(X, Y, theta):
mu = np.array([SCALE * theta.mu_alpha + theta.mu_zeta for _ in Y])
return mu
sim.mean = mean
def cov_old(X, Y, theta):
A = theta.sig_alpha * np.dot(
np.array(X)[:, 1:-3],
np.array(X)[:, 1:-3].T)
B = theta.sig_beta * np.diag(np.ones(len(X)))
C = theta.sig_zeta
D = mk52(np.array(X)[:, -3:-1], [theta.l1, theta.l2], theta.sig_m)
return theta.rho_matrix(X) * (A + B + C + D)
def cov_old2(X, Y, theta):
A = theta.sig_alpha * np.dot(
np.array(X)[:, 1:-3],
np.array(X)[:, 1:-3].T)
B = theta.sig_beta * np.diag(np.ones(len(X)))
C = theta.sig_zeta
D = mk52(np.array(X)[:, -3:-1], [theta.l1, theta.l2], theta.sig_m)
return theta.rho_matrix(X, use_psd=True) * (A + B + C + D)
def cov_new(X, Y, theta):
# Get a list of all unique X, removing initial IS identifier
X0 = []
for x in X:
if not any(
[all([a == b for a, b in zip(x[1:], xchk)]) for xchk in X0]):
X0.append(x[1:])
A = theta.sig_alpha * np.dot(
np.array(X0)[:, :-3],
np.array(X0)[:, :-3].T)
B = theta.sig_beta * np.diag(np.ones(len(X0)))
C = theta.sig_zeta
D = mk52(np.array(X0)[:, -3:-1], [theta.l1, theta.l2], theta.sig_m)
Kx = A + B + C + D
L = np.array([
np.array([
theta.rho[str(sorted([i, j]))] if i >= j else 0.0
for j in range(theta.n_IS)
]) for i in range(theta.n_IS)
])
# Normalize L to stop over-scaling values small
L = L / np.linalg.norm(L)
# Force it to be positive semi-definite
Ks = L.dot(L.T)
return np.kron(Ks, Kx)
#K = np.kron(Ks, Kx)
# Now, we get the sub-covariance matrix for the specified sampled X and Y
indices = []
for l in range(theta.n_IS):
for i, x in enumerate(X0):
test = [l] + list(x)
if any(
[all([a == b for a, b in zip(test, xchk)]) for xchk in X]):
indices.append(l * len(X0) + i)
K_local = K[np.ix_(indices, indices)]
return K_local
sim.cov = cov_new
sim.theta.bounds = {}
sim.theta.mu_alpha, sim.theta.bounds['mu_alpha'] = None, (
1E-3, lambda _, Y: max(Y))
sim.theta.sig_alpha, sim.theta.bounds['sig_alpha'] = None, (
1E-2, lambda _, Y: 10.0 * np.var(Y))
sim.theta.sig_beta, sim.theta.bounds['sig_beta'] = None, (
1E-2, lambda _, Y: 10.0 * np.var(Y))
sim.theta.mu_zeta, sim.theta.bounds['mu_zeta'] = None, (
1E-3, lambda _, Y: max(Y))
sim.theta.sig_zeta, sim.theta.bounds['sig_zeta'] = None, (
1E-2, lambda _, Y: 10.0 * np.var(Y))
sim.theta.sig_m, sim.theta.bounds['sig_m'] = None, (1E-2,
lambda _, Y: np.var(Y))
sim.theta.l1, sim.theta.bounds['l1'] = None, (1E-1, 1)
sim.theta.l2, sim.theta.bounds['l2'] = None, (1E-1, 1)
# # NOTE! This is a reserved keyword in misoKG. We will generate a list of the same length
# # of the information sources, and use this for scaling our IS.
# sim.theta.rho = {"[0, 0]": 1, "[0, 1]": None, "[1, 1]": 1}
# sim.theta.bounds['rho [0, 1]'] = (-1.0, 1.0)
# sim.theta.bounds['rho [0, 0]'] = (1, 1)
# sim.theta.bounds['rho [1, 1]'] = (1, 1)
sim.theta.rho = {"[0, 0]": None, "[0, 1]": None, "[1, 1]": None}
sim.theta.bounds['rho [0, 0]'] = (0.1, 1.0)
sim.theta.bounds['rho [0, 1]'] = (0.1, 1.0)
sim.theta.bounds['rho [1, 1]'] = (0.1, 1.0)
sim.theta.set_hp_names()
sim.primary_rho_opt = False
#sim.update_hp_only_with_IS0 = True
sim.update_hp_only_with_overlapped = True
###################################################################################################
# Start simulation
sim.run()